Biochemistry 1995, 34, 3561-3568
3561
Steady-State Kinetics of the Binding of ,&Lactams and Penicilloates to the Second Binding Site of the Enterobacter cloacae P99 ,8-Lactamaset Marek Dryjanski and R. F. Pratt* Department of Chemistry, Wesleyan University, Middletown, Connecticut 06459 Received May 31, 1994; Revised Manuscript Received December 7, 1994@
ABSTRACT: Previous research has shown that the class C p-lactamase of Enterobacter cloacae P99 is able to catalyze the hydrolysis and aminolysis of acyclic depsipeptides. The steady kinetics of these reactions are complicated by the presence of an additional (depsi)peptide binding site in addition to the active site [Pazhanisamy, S., & Pratt, R. F. (1989) Biochemistry 28, 6875-68821. The present paper presents a steady-state kinetic analysis of the inhibition of depsipeptide hydrolysis by sodium benzylpenicilloate, methyl benzylpenicilloate, 6-aminopenicillanic acid, and 7-aminocephalospoianic acid. The two p-lactams are considerably poorer substrates than the depsipeptide employed, rn-[[(phenylacetyl)glycyl]oxy]benzoic acid. The aim was to determine the relative affinity of these ligands for the active site and the second site. Three types of experiments were employed: (i) measurements of direct inhibition of depsipeptide hydrolysis, (ii) measurements of the effect of an active-site-directed inhibitor, m-(dansylamidopheny1)boronic acid, on the effectiveness of the ligands as inhibitors, and (iii) measurements of the effect of a preferential second site ligand, N-(phenylacety1)glycyl-D-phenylalanine,on the effectiveness of the ligands as inhibitors. The results suggest that all four ligands preferentially bind to the active site, with weaker binding at the second site. The necessarily weaker binding of a ligand to the second site when the active site is occupied by a transition-state analog inhibitor was analyzed. Perhaps surprisingly, the intact p-lactams appeared to bind more firmly to the alternative site than do the flexible penicilloates. The results also show that ligands can be present'in the second site through all stages of depsipeptide and p-lactam hydrolysis. It is likely that the second site (or sites) represents the remnants of substrate binding subsites present on the evolutionary progenitors of p-lactamases, the bacterial DD-peptidases.
P-Lactamases continue to represent an important source of bacterial resistance to p-lactam antibiotics (Neu, 1992). As part of a program of exploration of the chemical properties of the active sites of these enzymes, we have demonstrated that they catalyze not only the hydrolysis of p-lactams but also the hydrolysis and aminolysis of acyclic depsipeptides of general structure 1 (Pratt & Govardhan, RCoNH
1
OAOR'(Ar, 1
1984; Govardhan & Pratt, 1987). The steady-state kinetics of these depsipeptidereactions are complex, as demonstrated by our detailed study of the hydrolysis and aminolysis of 2 catalyzed by the class C P-lactamase of Enterobacter cloacae P99 (Govardhan & Pratt, 1987; Pazhanisamy et al., 1989; Pazhanisamy & Pratt, 1989b). This complexity arises from the presence of a binding site on the enzyme, distinct from the active site, capable of binding acyclic peptides and depsipeptides. Its presence is most directly appreciated by the demonstration that 3, the product of aminolysis of 2 by D-phenylalanine, is a noncompetitive inhibitor of the hydrolysis of 2. The specificity of the aminolysis reaction with respect to the structure of the amine acceptor of the acyl group is indicative of an acceptor binding site, although This research was supported by the National Institutes of Health Grant AI-17986 to R.F.P. Abstract published in Advance ACSAbsrracrs, February 15, 1995. @
PhCHZCONH
PhCHpCONH
LTPh
0
OLOQ
cop-
con-
3
2
binding is weak (Pazhanisamy & Pratt, 1989a,b). The aminolysis kinetics indicate a competitive interaction between the second reactant/product binding site and the acceptor site, but the extent of physical overlap between the sites is not known. Thus, there may be one or two small molecule binding sites in addition to the active site where one of them, the acceptor site, must necessarily be directly adjacent to the productive binding site of a depsipeptide substrate. The existence of an extended binding site in the P99 p-lactamase most likely derives from the past history of P-lactamases and DD-peptidases. Functional and structural studies now strongly support the proposition of Tipper and Strominger (1965) that P-lactamases are the evolutionary descendants of the DD-peptidases that catalyze peptidoglycan cross-linking in the final step of bacterial cell wall biosynthesis. Indeed, it seems likely that the P99 P-lactamase, a typical class C enzyme, is structurally very similar to the Streptomyces R61 DD-peptidase (Ghuysen, 1991; Lobkovsky et al., 1993). In view of these considerations, it seems possible, if not likely, that the second binding site detected on the P99 p-lactamase derives from the extended binding site likely present on DD-peptidases to accommodate their polymeric peptidoglycan substrates. Indeed, kinetic studies of the R61 DD-peptidase indicated the productive binding of extended acyl donors and acceptors (Ghuysen et al., 1979;
0006-2960/95/0434-3561$09.00/0 0 1995 American Chemical Society
3562 Biochemistry, Vol. 34, No. 11, 1995
Dryjanski and Pratt
Frbre & Joris, 1985). Peptide amines have been shown to inhibit the aminolysis reaction in a way that was best interpreted in terms of a quaternary complex with one acyl donor and two amine acceptors bound to the enzyme (Frbre et al., 1973; Perkins et al., 1973; Ghuysen et al., 1974). Initial investigations of the catalystic activity of high molecular weight DD-peptidases, e.g., penicillin-binding protein l a of Escherichia coli, suggested preference for extended disaccharide pentapeptide substrates (Ishino et al., 1980). Further, recent studies of the R61 DD-peptidase with thiol depsipeptide substrates show, by similar steady-state kinetics of aminolysis to those observed with the P99 P-lactamase, the presence of an additional substrate binding site in that enzyme also (Jamin et al., 1993). It seems likely then that the DD-peptidases and therefore, by default, the B-lactamases have a complicated system of binding subsites around the residues encompassing the scissile bond at the active site. No structural analysis of these features has yet been attempted. We have been interested in these sites as possible targets for extended inhibitor design and, with this in mind, have been exploring the specificity with respect to ligand structure of the second binding site of the P99 P-lactamase. One interesting and important aspect of this inquiry was the question as to whether p-lactams or their hydrolysis products bind at the secondary site. In this paper, we present the results of a steady-state kinetic approach to this question. The experiments involved reaction velocity measurements in the presence of a substrate, the depsipeptide 2, and the potential ligand, either benzylpenicilloate (4), methyl benzylpenicilloate (5), 6-aminopenicillanic acid (6), or 7-aminocephalosporanic acid (7). In order to clarify the mode of binding of these ligands, experiments in the presence of a competitive inhibitor, 3-(dansy1amido)phenylboronic acid (S), and separately a noncompetitive inhibitor, N-(phenylacety1)glycyl-D-phenylalanine (3) were also performed. The kinetics were complicated by the ability of the substrate to also bind at the secondary site. The results suggest that 4-7 all bind to some extent to the secondary site. Peptides are also able to bind there when the active site is occupied by a p-lactam. P h C H z CH OI N! H n -
P h C H & O HN HI n! >
Cop- NH
Me02C
NH
Coz-
CO*4
5
H2N
0
EXPERIMENTAL PROCEDURES Materials. Monosodium benzylpenicilloate (4), mp 160162 "C, was prepared by alkaline hydrolysis of benzylpeni-
cillin (Sigma) and recrystallized from water, as described by Mozingo and Folkers (1949). Material with the same 'H NMR spectrum is obtained as the immediate product from /?-lactamase-catalyzed hydrolysis of benzylpenicillin, and thus the isolated alkaline hydrolysis product presumably has the "natural" 5R,6R configuration (4). Methyl benzylpenicilloate, also of the 5R,6R configuration (5), was prepared by methanolysis of benzylpenicillin as described by Busson et al. (1976); after recrystallization from methanol-diethyl ether (1:9), the product had a mp of 126-127 "C. 6-Aminopenicillanic acid was purchased from Sigma Chemical Co. and used as received. 7-Aminocephalosporanic acid was obtained from Eli Lilly and Co. as a gift. The substrate, m-[[(phenylacetyl)glycyl]oxy]benzoic acid (2) was prepared as previously described (Govardhan & Pratt, 1987). m(Dansylamidopheny1)boronic acid was available from a previous synthesis in these laboratories (Pazhanisamy & Pratt, 1989b) or purchased from Sigma Chemical Co.; the two preparations had identical inhibitory properties. The peptide 3 was also available from a previous synthesis (Pazhanisamy & Pratt, 1989b). The P-lactamase of E. cloacae P99 was obtained from the Centre for Applied Microbiology and Research (Porton Down, U.K.) and, as previously (Govardhan & Pratt, 1987; Pazhanisamy & Pratt, 1989b), used as supplied. Kinetic Methods. The rates of hydrolysis of 2 in the presence of 3-8 were monitored spectrophotometrically by means of a Perkin-Elmer Lambda 4B spectrophotometer at 300 nm. All reactions were carried out in 20 mM MOPS buffer at pH 7.5 and at 25.0 "C. In all experiments, appropriate volumes of buffered stock solutions of 2-8 were mixed in a cuvette with buffer to a total volume of 0.8 mL. After temperature equilibration in the spectrophotometer had been achieved, the reaction was initiated by the addition of a 10-pL aliquot of a stock enzyme solution, giving a final enzyme concentration of ca. 35 nM; the P-lactamase concentrations of stock solutions were determined spectrophotometrically (Pazhanisamy et al., 1989). From the measurements of absorption vs time, initial rates of reaction were determined or, at substrate concentrations, less than 0.05 mM. (K, = 0.4 rnM), pseudo-first order rate constants were determined by nonlinear least-squares curve fitting (Johnson et al., 1976). Three types of experiments were performed: (i) Fixed substrate (D) concentration (0.02- 10 mM) and variable test ligand (I) concentrations: 4, 0-38; 5, 0-20; 6, 0-15; 7, 0-2 mM. (ii) Fixed D ( 3'
Kn 1Y
0.28 f 0.02 1.4 f 0.1 1.5 f 0.3 1.36 f 0.2 0.16 f 0.02 0.15 f 0.02 0.13 f 0.02 0.06 f 0.01 0.053 f 0.006 0.06 f 0.01
KW
> 15'
1.4 f 0.3 3.5
f 1.0
17.5 f 5
8 f 4 2.6
0.082 f 0.005
f 0.3
0.41 f 0.03
0.8 f 0.2 0.15 f 0.02 0.06 f 0.01
0.29 f 0.03 0.31 f 0.10 0.06 & 0.01
KI and U I refer to Scheme 2 (major binding of I to site I), and their alternatives, 61 and ~ K ' Irefer to Scheme 3 (major binding of I to site 2); Kn and KW refer to Schemes 4 and 5, respectively. J represents dansylboronate (S),and N is the peptide (3). The uncertainties quoted are standard deviations. Calculated with the assumption that KJ (Scheme 4) = 0.6 pM. Calculated with the assumption that KN (Scheme 5 ) = 1.0 mM. Calculated with the assumption that K1 (Schemes 2 and 3) = 1.7 mM. e '90% confidence. f Uncertain limits; '90% confidence that KJI> 3 mM. g 290% confidence that KJI > 4 mM.
3566 Biochemistry, Vol. 34, No. 11, 1995
Dryjanski and Pratt
Scheme 7 I
EIlN2
E11
Scheme 8 kat
D
E -ED, Km
-+ v J
I
=ED,*
Ill/oI
IllKl
E
IIlK,,*
ED112 -ED1*12
E12 YKm
P
E12
+
P
Pkat
that of 4 where, if it is assumed that KI (Scheme 7) is 1.4 mM (i.e., assuming no interaction between I in site 1 and N in site 2), a value of K'[ of 0.56 mM can be calculated from the apparent dissociation constant KI of Table 1 (taking an average value of 0.4 mM for the calculation). The assumption of no interaction between I and N when bound together may well not be a good one however. The binding of 3 to site 2 is weakend 2.8-fold on the binding of D to site 1 for example (Pazhanisamy & Pratt, 1989b). Experiment ii, discussed below, bears further on this point. The binding of I to EJ, measured in experiment ii, is clearly considerably weaker than to E. Since J, the boronate 8, likely to be a transition state analog (Beesley et al., 1988; Baldwin et al., 1991), is certainly bound in site 1, the implication is clear, in general agreement with that of experiment iii, that the preferential binding of 4-7 is to site 1. Although this conclusion is likely to be correct, it is worth reflecting on one point relevant to the interpretation of experiment ii, the fact, just mentioned, that 8 is a transitionstate analog inhibitor. The significance of this with respect to site 2 binding is seen by inspection of Scheme 8. This includes an equilibrium between the transition states EDI' and EDl'I? which, in reality of course, cannot be established. Nonetheless, on the basis of transition-state theory, a thermodynamic cycle involving the transition states can be employed (Wolfenden, 1975) from which eq 9 follows:
K': = (y/p)K',
(9)
If (yip) > 1, K'I~> K'I, I binds more strongly to E than to EDl', and I is an inhibitor at low [D]. Conversely, if (y/j3) < 1, K'3, < K'I, I binds more tightly to EDl' than to E, and I is an activator at low [D]. To take a known example in this system, that of 3, previous results show it to be a weak inhibitor at low [D] (Pazhanisamy & Pratt, 1989b). In accord with this, experiment yielded y = 2.8 and ,!? = 1.7; thus ( y / p ) = 1.65 and hence K'I' > K'I. A similar analysis should apply when the substrate D is replaced by a transition-state analog At. An inhibitor I (yl,!? 1) binding in site 2 will bind more tightly to E than to EA', and conversely an activator will bind more tightly to EA'. For a good transition-state analog, the ratio (yl,!?) obtained from studies of the binding of I to E and EA* should be the same as that from those of the effects of I on turnover of the substrate D analogous to At. This may be a useful criterion of the quality of a transition-state analog and one
which could be applied to both noncovalently and covalently (Rahil & Pratt, 1994) bound transition-state analogs. The essential requirement for such a test would be an effector binding site linked to the active site. This is restrictive, but a commonly available possibility is a proton binding site. With respect to the data of this paper, the implication of the above analysis is that 4-7, all of which are inhibitors at low [D], will bind more weakly to site 2 of EJI than they would to EDI, because J (8) is a transition-state analog. To test this hypothesis further, we performed experiment ii employing 3 (0-10 mM) rather than 4-7. The results (not shown) indicated KJIx 2 K'I, Le., that 3 bound more tightly to the free enzyme than to the transition-state analog complex EJ, as would be expected of an inhibitor. This result (y/,!? RZ 2) is in acceptable agreement with that quoted above for 3 as an inhibitor of D hydrolysis. Hence the binding of 4-7 to site 2, if site 1 were not occupied by a transition-state analog, would be perhaps twice as tight as suggested by Kn of Table 1. Although the binding of these ligands to site 1 would then still be tighter than to site 2, the difference would not be as great as it now appears. If, for example, K J were ~ 0.4 mM for the binding of 6 to EJ', where J' was a ground-state analog, then the binding of 6 to site 2 would not be greatly weaker, ca. 3-fold, than to site 1; the same conclusion would apply to 7. In fact, if it were somewhat tighter, the observation of a small value of j3 would follow. The conclusion of substantial, but not absolute, preferences of 4-7 for site 1 is supported by the data on the binding of I to ED. Previous results (Pazhanisamy & Pratt, 1989b) showed that the binding of D to Site 1 of E was essentially identically as strong as its binding to ED?. One might anticipate therefore that if I bound preferentially to site 1, its binding to E would be of comparable strength of that to ED*. With the exception of 4 (see below), this does seem to be true-KI is similar in magnitude to aKI. On the other hand, the affinity of D for E, binding in site 2, is also very similar to its affinity for ED1 (Pazhanisamy & Pratt, 1989b). One might therefore anticipate, if I preferably bound to site 2, that the binding of I to E would be comparably strong to that of I with ED1. This is certainly less true than the alternative above-K'I is smaller than yK'1 for all 4-7. Thus, the data seem best interpreted in terms of the preferential binding of 4-7 to site 1. Weaker but significant binding to site 2 is also observed, particularly for the p-lactams 6 and 7 as seen most clearly in the KJI values (Table 1). It is now informative to consider these conclusions in terms of the structures of 2-8. First, it is presumably unsurprising to find that the preferential binding of the p-lactams 6 and 7 is to site 1, the active site, since these compounds are substrates of this enzyme. The inhibition constants of these compounds, KI, should therefore be equal to the K, values for their turnover. The KI value for 6, 0.15 f 0.02 mM, is not unreasonably different from the directly determined Km,0.29 f 0.05 mM, but that for 7, 0.06 f 0.01 mM, seems distinctly different from its K,, 0.29 f 0.04 mM. The reason for these differences are not immediately clear but may well be the same as those responsible for a variety of substrate-specific hysteretic events that are not infrequently observed with P-lactamases (Zyk & Citri, 1968; Citri et al., 1976; Hashizume et al., 1988; Monks & Waley, 1988) although not yet
Second Binding Site of P99 P-Lactamase satisfactorily explained in molecular terms. In the present case, since the common substrate 2 is involved in all experiments, the qualitative interpretation of the results should be unaffected. The deacylation step is most likely rate-determining to the turnover of 6 and 7 at saturation (Knott-Hunziker et al., 1982; Govardhan & Pratt, 1987; Mazzella & Pratt, 1989; Monnaie et al., 1992), and thus Km-l probably represents binding to the stage of the acyl enzyme. That similar dissociation constants are obtained for 6 and 7 in the presence of 2 (a&) and 3 ( K I N indicates ) that (depsi)peptide species may bind to site 2 at the acyl enzyme stage of turnover. The presence of these ligands, at least, has little influence on acyl enzyme formation. Jamin et al. (1993) have concluded that depsipeptides may also bind to a second site on acyl enzymes formed on reaction between the same depsipeptide and the Streptomyces R6 1 DD-peptidase. This is likely true for the P99 P-lactamase also (Mazzella et al., 1991). The penicilloates 4 and 5 both inhibit the P99 P-lactamase, with the dianionic dicarboxylate 4 being somewhat more effective in binding to the free enzyme (lower KI). This result contrasts strikingly with that observed with the class A p-lactamase I of Bacillus cereus where 5 is by far the better inhibitor (Jones et al., 1989). The analogous penilloate, lacking the P-lactam-derived carboxyl group, is also a better inhibitor of this enzyme than is 4 (Kiener & Waley, 1978). Similarly perhaps, substrates of the class A enzyme having additional negative charge adjacent to the p-lactam ring, such as carbenicillin and sulbenicillin, are poorer substrates than their neutral analogs (Hardy et al., 1984). These observations with the class A enzyme have been interpreted in terms of unfavorable electrostatic interaction between the additional carboxylate and a negative charge in the active site, proposed to be Glu 166 by Jones et al. (1989). An alternative explanation would be a hydrophobic environment, although the crystal structures certainly indicate a rather polar active site. The present results with 4 and 5 could be incorporated into the electrostatic model in that class C enzymes do not have a carboxylate group in an analogous position to that of Glu 166 of the class A (Lobkovsky et al., 1993). They do have what may be a negatively charged Tyr 150 side chain, but the position of the anion would be different (Oefner et al., 1990; Lobkovsky et al., 1994). Thus, perhaps more under the influence of a local positive charge, e.g. of Lys 67, 4 may be the better inhibitor of the class C P99 P-lactamase. It is interesting that the presence of 2 or 3 in site 2 weakens the binding of 4 more than that of 5; this could represent unfavorable electrostatic interaction between the negatively charged ligands. Perhaps the most interesting result is the finding that 4-7 bind to site 2 when the active site is occupied by 8, a transition-state analog (&I). This makes it likely that such binding would also occur when a substrate undergoing turnover occupied the active site. As discussed above, this binding would probably be weakest at the transition state or, equivalently, to a transition-state analog. Certainly 2 and 3, as also discussed above, seem to be able to occupy site 2 during substrate turnover. Extrapolation of the latter result to the flexible 4 and 5 would not perhaps be surprising, but the suggestion that 6 and 7, compact p-lactam structures, bind so well to site 2 is certainly interesting particularly in view of the fact that their binding to site 1 involves acyl enzyme formation. Whether any covalent chemistry might
Biochemistry, Vol. 34, No. 11, 1995 3567
be involved in the interaction of 6 and 7 with site 2 is not known at present. It may be significant, in this regard at least, that the ED112 complexes where I is 6 or 7 have little or no ability to regenerate free enzyme, i.e., P (Scheme 8)
